From: Yonghong Song <yonghong.song@linux.dev>
To: Anton Protopopov <aspsk@isovalent.com>,
bpf@vger.kernel.org, Andrii Nakryiko <andrii@kernel.org>,
Daniel Borkmann <daniel@iogearbox.net>,
Eduard Zingerman <eddyz87@gmail.com>,
Quentin Monnet <qmo@kernel.org>,
Alexei Starovoitov <ast@kernel.org>
Subject: Re: [RFC PATCH bpf-next 00/14] instruction sets and static keys
Date: Tue, 18 Mar 2025 14:00:37 -0700 [thread overview]
Message-ID: <4a128a09-0b8b-488a-986b-7882f96bc5bb@linux.dev> (raw)
In-Reply-To: <20250318143318.656785-1-aspsk@isovalent.com>
On 3/18/25 7:33 AM, Anton Protopopov wrote:
> This patchset implements new type of map, instruction set, and uses
> it to build support for BPF static keys. The same map will be later
> used to provide support for indirect jumps and indirect calls. See
> [1], [2] for more context.
>
> Short table of contents:
>
> * patches 1, 9, 10, 11 are simple fixes (which can be sent
> independently, if acked)
>
> * patches 2, 3 add a new map type, BPF_MAP_TYPE_INSN_SET, and
> corresponding selftests. This map is used to track how original
> instructions were relocated into 'xlated' during the verification
>
> * patches 4, 5, 6, 7, 8 add support for static keys (kernel only)
> using (an extension) to that new map type. Only x86 support is
> added in this RFC
>
> * patches 12, 13, 14 add libbpf-side support for static keys and
> selftests
>
> It is RFC for a few reasons:
>
> 1) The kernel side of the static keys looks clear, however, the
> libbpf side is not _that_ clear. I thought that this is better to
> commit to a particular userspace design, as any particular design
> requires a lot of changes on the libbpf side. See patch 12 for
> the details
>
> 2) The libbpf part of the series requires a patched LLVM (see [3]),
> which adds support for gotol_or_nop/nop_or_gotol instructions, so
> selftests would not compile in CI.
>
> 3) Patch 4 adds support for a new BPF instruction. It looks
> reasonable to use an extended BPF_JMP|BPF_JA instruction, and not
> may_goto. Reasons: a follow up will add support for
> BPF_JMP|BPF_JA|BPF_X (indirect jumps), which also utilizes INSN_SET maps (see [2]).
> Then another follow up will add support CALL|BPF_X, for which there's
> no corresponding magic instruction (to be discussed at the following
> LSF/MM/BPF).
>
> Besides these reasons, there are some questions / known bugs,
> which will be fixed once the general plan is confirmed:
>
> * bpf_jit_blind_constants will patch code, which is ignored in this
> RFC series. The solution would be either moving tracking
> instruction sets to bpf_prog from the verifier environment,
> or moving bpf_jit_blind_constants upper the stack (right now,
> this is the first thing which every jit does, so maybe it can
> be actually executed from the verifier, and provide env context)
>
> * gen-loader not supported, fd_array usage in libbpf should be
> re-designed (see patch 12 for more details)
>
> * insn_off -> insn_set map mapping should be optimized (now it is
> brute force)
>
> Links:
> 1. http://oldvger.kernel.org/bpfconf2024_material/bpf_static_keys.pdf
> 2. https://lpc.events/event/18/contributions/1941/
> 3. https://github.com/aspsk/llvm-project/tree/static-keys
For llvm patch in [3], please remove changes in function isValidIdInMiddle()
as gotol_or_nop or nop_or_gotol will not appear in the *middle* of any
instruction. "gotol" should not be there either, I may remove it sometime
later.
>
> Anton Protopopov (14):
> bpf: fix a comment describing bpf_attr
> bpf: add new map type: instructions set
> selftests/bpf: add selftests for new insn_set map
> bpf: add support for an extended JA instruction
> bpf: Add kernel/bpftool asm support for new instructions
> bpf: add BPF_STATIC_KEY_UPDATE syscall
> bpf: save the start of functions in bpf_prog_aux
> bpf, x86: implement static key support
> selftests/bpf: add guard macros around likely/unlikely
> libbpf: add likely/unlikely macros
> selftests/bpf: remove likely/unlikely definitions
> libbpf: BPF Static Keys support
> libbpf: Add bpf_static_key_update() API
> selftests/bpf: Add tests for BPF static calls
>
> arch/x86/net/bpf_jit_comp.c | 65 +-
> include/linux/bpf.h | 28 +
> include/linux/bpf_types.h | 1 +
> include/linux/bpf_verifier.h | 2 +
> include/uapi/linux/bpf.h | 40 +-
> kernel/bpf/Makefile | 2 +-
> kernel/bpf/bpf_insn_set.c | 400 +++++++++++
> kernel/bpf/core.c | 5 +
> kernel/bpf/disasm.c | 33 +-
> kernel/bpf/syscall.c | 28 +
> kernel/bpf/verifier.c | 94 ++-
> tools/include/uapi/linux/bpf.h | 40 +-
> tools/lib/bpf/bpf.c | 17 +
> tools/lib/bpf/bpf.h | 19 +
> tools/lib/bpf/bpf_helpers.h | 63 ++
> tools/lib/bpf/libbpf.c | 362 +++++++++-
> tools/lib/bpf/libbpf.map | 1 +
> tools/lib/bpf/libbpf_internal.h | 3 +
> tools/lib/bpf/linker.c | 6 +-
> .../selftests/bpf/bpf_arena_spin_lock.h | 3 -
> .../selftests/bpf/prog_tests/bpf_insn_set.c | 639 ++++++++++++++++++
> .../bpf/prog_tests/bpf_static_keys.c | 359 ++++++++++
> .../selftests/bpf/progs/bpf_static_keys.c | 131 ++++
> tools/testing/selftests/bpf/progs/iters.c | 2 -
> 24 files changed, 2315 insertions(+), 28 deletions(-)
> create mode 100644 kernel/bpf/bpf_insn_set.c
> create mode 100644 tools/testing/selftests/bpf/prog_tests/bpf_insn_set.c
> create mode 100644 tools/testing/selftests/bpf/prog_tests/bpf_static_keys.c
> create mode 100644 tools/testing/selftests/bpf/progs/bpf_static_keys.c
>
next prev parent reply other threads:[~2025-03-18 21:00 UTC|newest]
Thread overview: 31+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-03-18 14:33 [RFC PATCH bpf-next 00/14] instruction sets and static keys Anton Protopopov
2025-03-18 14:33 ` [RFC PATCH bpf-next 01/14] bpf: fix a comment describing bpf_attr Anton Protopopov
2025-03-18 14:33 ` [RFC PATCH bpf-next 02/14] bpf: add new map type: instructions set Anton Protopopov
2025-03-20 7:56 ` Leon Hwang
2025-03-20 9:34 ` Anton Protopopov
2025-03-18 14:33 ` [RFC PATCH bpf-next 03/14] selftests/bpf: add selftests for new insn_set map Anton Protopopov
2025-03-18 20:56 ` Yonghong Song
2025-03-19 17:26 ` Anton Protopopov
2025-03-19 17:30 ` Anton Protopopov
2025-03-18 14:33 ` [RFC PATCH bpf-next 04/14] bpf: add support for an extended JA instruction Anton Protopopov
2025-03-18 19:00 ` David Faust
2025-03-18 19:24 ` Anton Protopopov
2025-03-18 19:30 ` David Faust
2025-03-18 19:47 ` Anton Protopopov
2025-03-18 14:33 ` [RFC PATCH bpf-next 05/14] bpf: Add kernel/bpftool asm support for new instructions Anton Protopopov
2025-03-18 14:33 ` [RFC PATCH bpf-next 06/14] bpf: add BPF_STATIC_KEY_UPDATE syscall Anton Protopopov
2025-03-18 14:33 ` [RFC PATCH bpf-next 07/14] bpf: save the start of functions in bpf_prog_aux Anton Protopopov
2025-03-18 14:33 ` [RFC PATCH bpf-next 08/14] bpf, x86: implement static key support Anton Protopopov
2025-03-18 14:33 ` [RFC PATCH bpf-next 09/14] selftests/bpf: add guard macros around likely/unlikely Anton Protopopov
2025-03-18 14:33 ` [RFC PATCH bpf-next 10/14] libbpf: add likely/unlikely macros Anton Protopopov
2025-03-28 20:57 ` Andrii Nakryiko
2025-03-29 13:38 ` Anton Protopopov
2025-03-31 20:10 ` Andrii Nakryiko
2025-03-18 14:33 ` [RFC PATCH bpf-next 11/14] selftests/bpf: remove likely/unlikely definitions Anton Protopopov
2025-03-18 14:33 ` [RFC PATCH bpf-next 12/14] libbpf: BPF Static Keys support Anton Protopopov
2025-03-18 14:33 ` [RFC PATCH bpf-next 13/14] libbpf: Add bpf_static_key_update() API Anton Protopopov
2025-03-18 14:33 ` [RFC PATCH bpf-next 14/14] selftests/bpf: Add tests for BPF static calls Anton Protopopov
2025-03-18 20:53 ` Yonghong Song
2025-03-18 21:00 ` Anton Protopopov
2025-03-18 21:00 ` Yonghong Song [this message]
2025-03-19 17:45 ` [RFC PATCH bpf-next 00/14] instruction sets and static keys Anton Protopopov
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4a128a09-0b8b-488a-986b-7882f96bc5bb@linux.dev \
--to=yonghong.song@linux.dev \
--cc=andrii@kernel.org \
--cc=aspsk@isovalent.com \
--cc=ast@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=daniel@iogearbox.net \
--cc=eddyz87@gmail.com \
--cc=qmo@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox